Synthese

https://doi.org/10.1007/s11229-021-03163-y

ORIGINAL RESEARCH

A kind route from grounding to fundamentality

Fabrice Correia1 

Received: 17 September 2020 / Accepted: 16 April 2021

© The Author(s) 2021

Abstract
I offer an account of fundamentality for facts in terms of metaphysical grounding.
The account does justice to the idea that whether a fact is absolutely fundamental,
and whether a fact is more fundamental than, or as fundamental as, another fact,
are a matter of where in a grounding-induced hierarchy of kinds of facts these facts
appear.

Keywords  Fundamentality · Relative fundamentality · Absolute fundamentality ·

Grounding

1 Introduction

The notion of (fact) fundamentality comes in two varieties, absolute and relative.
Absolute fundamentality is at work when we say that a fact is fundamental tout
court. Relative fundamentality itself comes in two varieties: there is the notion of
being more fundamental than and that of being as fundamental as. It is natural to
think—and it actually has been thought—that (metaphysical) grounding can be used
to characterise a notion of fundamentality of each variety.1 Thus, many philosophers
have recently held the view that one way for a fact to be absolutely fundamental
is for it to be ungrounded.2 Two of these philosophers, Karen Bennett (2017) and

1
  In Correia (2021), I explore an opposite view, according to which it is grounding that is characterised
in terms of relative fundamentality. This view will not be discussed here. I should mention that at the
time of writing that paper I was a bit sceptical about the possibility of formulating an account of relative
fundamentality of the sort I put forward in the present paper (see pp. 1280–1281).
2
  See for instance Audi (2012: p. 710), Bennett (2011: p. 1), Bennett (2017: p. 106), Dixon (2016: p.
442), Leuenberger (2020: p. 2650), Rosen (2010: p. 112), Schaffer (2009: p. 373), Shumener (2017: p.
2) and Wallner (2021: p. 1260).

* Fabrice Correia
fabrice.correia@unige.ch
1
Department of Philosophy, University of Geneva, Rue De‑Candolle 5, 1211 Geneva 4,
Switzerland

13
Vol.:(0123456789)
 Synthese

Jonas Werner (forthcoming), have proposed characterisations of relative fundamen-

tality in terms of grounding, and I did the same in Correia forthcoming.3
Werner rightly points out that Bennett’s characterisation does not capture a uni-
fied notion. At least part of the reason is that it merges two quite different ideas: (i)
the idea that the fundamentality status of a fact—that is, whether it is fundamental
tout court, and whether it is more fundamental than, or as fundamental as, other
facts—is a matter of how far, ground-theoretically speaking, this fact is from the
ungrounded facts, and (ii) the idea that the fundamentality status of a fact is a mat-
ter of where, in a grounding-induced hierarchy of kinds of facts, this fact appears.4
Werner sets himself the task of characterising a notion of fundamentality that does
justice to the first idea, and I also do it in Correia forthcoming (I actually character-
ise two such notions there). In the present paper, I set myself the task of characteris-
ing a notion of fundamentality that does justice to the second idea.5
The plan is as follows. Section  2 deals with preliminary considerations about
grounding and relative fundamentality. In Sect. 3, I present Bennett’s account of the
relation of being more fundamental than (hereafter: <–fundamentality), of the rela-
tion of being as fundamental as (hereafter: = –fundamentality) and of absolute fun-
damentality in terms of grounding, and I give some reasons to reject her account
of the first relation. In Sect.  4, I clarify a bit the conception of fundamentality I
am interested in. In Sect. 5, I put forward my own account of <–fundamentality and
I argue that it is superior to Bennett’s. In Sect.  6, I deal with the sister accounts
of = –fundamentality and absolute fundamentality.

2 Preliminary considerations on grounding and fundamentality

I will assume that grounding has the following features:

(a) It is a relation, that is, it is expressed by means of a relational predicate, whose

relata are states of affairs or propositions. I will not make any assumption about
which of the two options is correct. It is understood that the relata of relative
fundamentality and the bearers of absolute fundamentality belong to the same
category as the relata of grounding, whichever it may be. I will use ‘facts’ as a
label for the relevant category of entities.

3
  Werner only implicitly identifies being fundamental with being ungrounded.
4
  I say ‘part of the reason’ partly because I am not opposed in principle to the view that these two ideas
may be combined into an account of a unified notion (although I have no idea how this could be done).
What I think creates dis-unity is, among other things, the specific way in which these two ideas are put
together. See Sect. 3 for the details of Bennett’s characterisation and more on the issue of unity.
5
  The idea that reality is hierarchically structured into various layers (or levels) and that entities are more
or less fundamental than, or as fundamental as, others depending on which layer they belong to, is admit-
tedly fairly widespread and predates the current massive interest in the notion of grounding (see Kim
(2002) for a nice discussion of this idea from the pre-grounding era). The idea that some structures of
this sort are generated by grounding (rather than, say, emergence, supervenience or mereological rela-
tions) is nowadays also contemplated by a number of philosophers (apart from Bennett (2017), see, inter
alia, deRosset (2013) and Rabin (2018)).

13
Synthese

(b) It is factive, that is, it relates only obtaining states of affairs or true propositions.
(c) It is many-one, that is, what is grounded is always one fact, and what grounds
something can be one or several facts taken together.
(d) It is transitive—to be more accurate, it obeys the following “cut principle”:6
For every indexed family of pluralities of facts (Δi)i∈I, every indexed family of
facts with the same index set (­Fi)i∈I, every plurality of facts Γ and every fact
G, if the ­Fis together with the members of Γ ground G, then provided that the
members of Δi ground ­Fi for all i ∈ I, the members of the Δis together with
the members of Γ also ground G.
I will also understand ‘partially grounds’ according to the following standard
definition:
Fact F partially grounds fact G ≡ G is grounded in F, or in a plurality of facts
that comprises F.7
Thus, ‘partial’ is used in a loose sense: full grounds count as partial grounds. Given
the definition, (d) above entails that partial grounding is transitive.
Assumption (a) is dispensable. The discussion could be run in much the same
way on the assumption that grounding is expressed by means of a sentential opera-
tor rather than by means of a predicate (and that the same holds of both relative and
absolute fundamentality).8 The reason for assuming the predicate mode of expres-
sion is just convenience: going for an operator regimentation of grounding would
force one to appeal to higher-order quantifiers where the predicate regimentation
only requires first-order quantifiers over facts, and first-order quantification is more
familiar than higher-order quantification.
Assumption (b) is also dispensable: I could have worked with a non-factive
notion of grounding throughout. However, given that relative fundamentality and
absolute fundamentality are themselves factive, working with a non-factive notion
of grounding would have forced me to often slightly complicate the formulation of
principles and characterisations by adding conditions to the effect that certain states
of affairs obtain / certain propositions are true. Note that since given a non-factive
notion of grounding, a corresponding factive notion is easily definable in terms of it,
assumption (b) is not ideologically demanding.9
By contrast, the discussion to come will substantially rely on assumptions (c) and
(d), at some points at least. The alternative to (c) is the very unorthodox view that
grounding is many-many: what does the grounding can be one or several facts taken

6
 The principle is formulated in this general form for instance by Litland (2013: p. 20). Transitiv-
ity proper is a property of one–one relations, satisfaction of the cut principle is the natural correlate for
many-one relations.
7
  Here and everywhere else below, I use the equivalence symbol ‘≡’ indistinctively to point to stipula-
tive definitions and to point to definitions / characterisations that are intended to capture previously given
notions. Context should always determine which meaning the symbol has.
8
  See Correia (2010) on the predicate vs operator regimentation.
9
  See Fine (2012) on the factive vs non-factive distinction.

13
 Synthese

together, and what is grounded can also be one or several facts taken together.10
Many-many grounding is far from being well understood, and it is for this reason
that I leave it aside. The view that grounding obeys the cut principle is hard to reject
and has indeed hardly ever been contested.11 Hence, having (d) on board is not a
very substantial move—or so I take it. That said, the fact that (c) and (d) are very
plausible and that they are part of grounding orthodoxy makes it perfectly appropri-
ate to take them on board. However, the discussion to come can certainly serve as a
useful basis for the study of the connections between grounding and fundamentality
on the assumption that (c) or (d) should be rejected.
Finally, let me stress that I will take as a constraint on a correct account of <–fun-
damentality and = –fundamentality that it should predict that the former relation is
both transitive and irreflexive (and hence asymmetric) and that the latter relation is
an equivalence relation, i.e., reflexive, symmetric and transitive. That <–fundamen-
tality and = –fundamentality have the respective features sounds pre-theoretically
obvious indeed.

3 Bennett on fundamentality

Bennett (2017) takes grounding to be but one member of a large class of “building
relations”, and she proposes a characterisation of <–fundamentality, = – fundamen-
tality and absolute fundamentality indexed to members of this class. Given my aims,
I will from now on pretend that she focuses on grounding.
Bennett proposes to characterise the three notions under focus as follows (I use
her labels for the first characterisation; throughout the paper, I use ‘F’, ‘G’ and
the like as variables for facts and ‘ K  ’, ‘ℒ’ and the likes for variable for kinds of
facts):12
(MFT) F is more fundamental than G ≡ at least one of the following holds:

(1) F is fewer grounding steps away from the ungrounded fact(s) that terminate
its unique chain than G is from the ungrounded fact(s) that terminate its
unique chain;
(2) F partially grounds G;
(3) F stands in the ancestral of partial grounding to G;13
(4) F is ungrounded while G is grounded;
(5) F belongs to some kind K and G belongs to some kind ℒ such that

10
  See Dasgupta (2014) and Litland (2016).
11
  Schaffer (2012) proposes counterexamples to the transitivity of partial grounding, but Litland (2013)
gives very convincing replies.
12
  See p. 161 for the first notion, p. 173 for the second one and ch. 5 for the third one. In clause (1) of
(MFT), ‘chain’ means chain of facts related by grounding. The idea that every fact has a unique chain
that terminates in ungrounded facts is problematic in more than one way. See Werner forthcoming and
Correia forthcoming for discussions.
13
  Of course, clauses (2) and (3) are equivalent if we assume, as I did right from the start, that partial
grounding is transitive.

13
Synthese

(a) neither K nor ℒ includes both grounded and ungrounded members, and
(b) G does not belong to K and F does not belong to ℒ, and
(c) ℒs are typically or normally grounded in K s.

(AFA) F is as fundamental as G ≡ for all facts F*, (i) F* is more fundamental

than F iff F* is more fundamental than G, and (ii) F is more fundamen-
tal than F* iff G is more fundamental than F*.
(F) F is fundamental ≡ F is ungrounded.

As I stressed in the introduction, (F) corresponds to a widespread conception of

absolute fundamentality. (AFA) is a very natural characterisation of = –fundamental-
ity if we suppose given <–fundamentality. Bennett credits Jon Litland with the idea
of this characterisation. Bennett’s substantial contribution here is (MFT).
If I were happy with (MFT), I would not have anything to say against (AFA) or
(F), and I would therefore be happy with the whole Bennettian package. However,
I find (MFT) unsatisfactory, for many reasons indeed. Let me here mention three.
(I have further objections which concern matters of detail, but they can be circum-
vented by appropriately modifying the account; see Correia forthcoming for objec-
tions linked to clause (1).)

I. (MFT) does not seem to capture, or even to be close to capturing, a unified

notion. One cannot help but have the feeling that in formulating (MFT), Ben-
nett tried to do justice to a bunch of different pre-theoretic intuitions that some-
times pull in quite different directions, sometimes even in opposite directions.
Werner (forthcoming) also highlights the disunity of the account, pointing
to the fact that while clauses (1)-(4) are purely ground-theoretic, clause (5)
crucially invokes the metaphysical notion of a kind. As it turns out, Bennett is
sympathetic to the view that our pre-theoretic conception of <–fundamental-
ity, taken as a whole, may be incoherent (2019: 332; see also 2017: 161). Yet
she seems to think that highly disjunctive characterisations in the vicinity of
(MFT)—if not (MFT) itself—can capture valuable notions. I doubt it. The
right thing to do, or so I believe, is to rather look for accounts of more unified
notions, even if this means that they may only very partially fit with our pre-
theoretic conception.
II. My second reason for dissatisfaction concerns the presence of the phrase ‘typi-
cally or normally’ in clause (5). Bennett does not say enough to fix its meaning.
She explains that in this context, normality should probably not be under-
stood in the standard, statistical sense, and she refers to Millikan (1989) for an
example of a non-statistical conception (Bennett, 2017: p. 159, fn. 20). Yet I
very much doubt that Millikan’s teleological notion of normality, whatever its
merits for theorising about biological functions, is the right notion to invoke
in the present context. And so, I am left wondering how clause (5) should be
understood.

13
 Synthese

III. The view that (5) is a sufficient condition for a fact F to be more fundamental
than a fact G yields unwanted results. Suppose there are the kind mental fact
and the kind biological fact. Suppose that mental facts are grounded in biologi-
cal facts, and that biological facts are grounded (say, in physical facts). Then
thanks to clause (5), (MFT) predicts that (A) every biological fact is more
fundamental than every mental fact. No problem yet. But suppose there are
the kind disjunction of conjunctions of atomic facts and the kind conjunction
of atomic facts and assume, as orthodoxy about the logic of grounding has
it,14 that every fact of the first kind is grounded in some fact(s) of the second
kind, and that every fact in the second kind is grounded in atomic facts. Then
thanks to clause (5), (MFT) predicts that (B) every conjunction of atomic
facts is more fundamental than every disjunction of conjunctions of atomic
facts. Now we have a problem. For let F be a mental fact that is a conjunction
of atomic mental facts, and let G be a biological fact that is a disjunction of
conjunctions of atomic biological facts. Then by (A), G is more fundamental
than F, and by (B), F is more fundamental than G. But since <–fundamentality
is asymmetric, this is impossible.

Some might protest against the previous argument, on the grounds that the kinds
disjunction of conjunctions of atomic facts and conjunction of atomic facts are artifi-
cial and that Bennett’s (5) should be understood as quantifying only over non-artifi-
cial kinds. I reply in three steps.
First, Bennett says almost nothing about which collections of facts (or objects
more generally) she takes to correspond to kinds. She is aware that some restrictions
are needed in the context of the formulation of clause (5) in (MFT), but the restric-
tions she identifies boil down to those expressed by subclauses (a) and (b) in clause
(5) (see Bennett, 2017: pp. 159–160). So far as I can see, nothing in what she says
rules out the possibility that there be kinds such as disjunction of conjunctions of
atomic facts and conjunction of atomic facts.
Second, it seems to me reasonable to hold that ‘conjunction of atomic sentences’
and ‘disjunction of conjunctions of atomic sentences’ are phrases that pick out non-
artificial collections of sentences. Such phrases are constantly used in logic text-
books, e.g., to pick out collections of sentences that are useful for certain purposes
(for defining conjunctive and disjunctive normal forms, for instance). If this is rea-
sonable, then it should be likewise reasonable to hold that the phrases ‘conjunction
of atomic facts’ and ‘disjunction of conjunctions of atomic facts’ pick out non-artifi-
cial collections of facts.
Third, let me add a further argument against Bennett’s (5) which follows the same
pattern. It invokes, not ground-theoretic connections stemming from logical form,
but ground-theoretic connections stemming from the determinable / determinate
relation, and it might for this reason be considered more convincing. Let me define
the following kinds:

• MACRO = the kind macrophysical fact

14
  See for instance Fine (2012: p. 58).

13
Synthese

• SWAN = the kind fact about swans

• DETERMINATE = the kind fact of type [x is F], where being F is a determinate
of being coloured
• DETERMINABLE = the kind fact of type [x is coloured]

Assume with grounding orthodoxy that every fact in DETERMINABLE is

grounded in some fact in DETERMINATE and assume that every fact about swans
is grounded in some macrophysical facts. Finally assume that none of these kinds
comprise both grounded and ungrounded facts. Then consider the fact [W is white],
where W is a regular swan, and the fact [C is coloured], where C is a given surface.
Bennett’s clause (5) predicts that [W is white] is more fundamental than [C is col-
oured] (due to the connection between DETERMINATE and DETERMINABLE)
and that the converse holds as well (due to the connection between MACRO and
SWAN). Hence, we have again a violation of the asymmetry of <–fundamentality.15

4 The kind‑theoretic conception of fundamentality

My first objection against (MFT) was that it merges different conceptions of <–fun-
damentality. In the introduction, I stressed that Bennett’s account of fundamentality
merges at least two different ideas, the idea that the fundamentality status of a fact
is a matter of how far, from a ground-theoretic point of view, the fact is from the
ungrounded facts, and the idea that the fundamentality status of a fact is a matter of
where, in a grounding-induced hierarchy of kinds of facts, the fact is located. The
first idea as applied to <–fundamentality is captured (if only imperfectly) in (MFT)
by clause (1), the second idea by clause (5). As I also stressed in the introduction,
my aim in this paper is to characterise a notion of fundamentality that meshes well
with the second idea. What is, more precisely, the conception of fundamentality I
am after—the kind-theoretic conception, as I will call it?
This conception involves the following two ideas:

• There is a relation—that I will call foundation—between kinds of facts which is

definable in ground-theoretic terms and which captures the idea that some kinds
of facts—the founded facts—metaphysically “arise from” some other kinds of
facts—the kinds that found them.

15
  In response to an objection against (MFT) put forward by Shumener (2019: pp. 5–6), Bennett (2019:
p. 332) suggests adding a subclause to (5) which says that the members of ℒ are same-ranked, where
two facts are same-ranked iff ‘they are the same number of [grounding] steps from the terminus of their
[grounding] chains’. (To be accurate, Shumener and Bennett formulate their points in terms of building
rather than grounding.) I ignored this modification, because it annihilates the possibility that clause (5)
secures, say, the view that biological facts are more fundamental than mental facts. For on the assump-
tion that every mental fact has a grounding chain with a terminus, the conjunction of two mental facts,
which is itself a mental fact, surely has a rank that is higher than the rank of any of these two facts.

13
 Synthese

• The fundamentality status of a fact—whether it is fundamental, and whether it is

more fundamental than, or as fundamental as, other facts—stems from the foun-
dational links between kinds of facts.

These ideas are neatly (although not adequately, see Sects. 3 and 5.1) illustrated by
clause (5) in Bennett’s (MFT): subclauses (a) and (c) jointly characterise a certain
sort of ground-theoretic connection between two kinds K and ℒ, and clause (5) as
a whole says that fact F is more fundamental than fact G when there are kinds K
and ℒ connected in the way in question such that F is in K but not in ℒ and G is in
ℒ but not in K .
A further important idea involved in the kind-theoretic conception of fundamen-
tality is the following:

• A fact may be more fundamental than another fact even though these two facts
are ground-theoretically completely disconnected.

To illustrate, consider the fact ­F0 that a given cell undergoes meiosis and the fact
­G0 that a given person experiences an acute pain, where the cell is in a remote gal-
axy and the person is somewhere on Earth.16 It makes perfect sense, on the kind-
theoretic conception of fundamentality, to claim that ­F0 is more fundamental than
­G0, even though (we may assume) ­F0 and ­G0 are not ground-theoretically connected.
This is also neatly illustrated by clause (5) in (MFT): if we assume that all mental
facts are grounded in biological facts, that all biological facts are grounded (say, in
physical facts), that ­F0 belongs to the kind biological fact but not to the kind mental
fact, and finally that ­G0 belongs to the latter kind but not to the former kind, then the
clause predicts that ­F0 is more fundamental than ­G0.
A few words about the notion of a kind of facts that is involved in the kind-the-
oretic conception of fundamentality are in order before moving on to the next sec-
tion. The sort of account that I intend to formulate arguably requires to impose some
restrictions on which collections of facts correspond to kinds of facts. This can be
illustrated with the view that clause (5) in (MFT) is sufficient for <–fundamentality.
Suppose there are the kind mental fact and the kind biological or mathematical fact.
Suppose that the mental facts are grounded in biological facts, and that the biologi-
cal facts and the mathematical facts are grounded (perhaps ultimately in divine facts,
or in physical facts and logical facts, respectively). Then if clause (5) in (MFT) is
sufficient for <–fundamentality, every mathematical fact is more fundamental than
every mental fact. Of course, this is something that one can hold. But this is cer-
tainly not something that one must hold, even if one accepts the assumptions that
were made.
This objection to the view that clause (5) is sufficient for <–fundamentality
will plausibly be blocked if we require that the kinds of facts appealed to in the
clause should be natural: it is indeed plausible to holds that the kind biological

16
  Bennett (2017: p. 149) gives a similar example to illustrate the fact that an entity may be more funda-
mental than another one even if the latter is not “built” from the former.

13
Synthese

or mathematical fact is non-natural, on any reasonable account of what a natural

kind of facts is. Similar objections are likely to be effective against any imple-
mentation of the kind-theoretic conception of fundamentality if non-natural kinds
of facts are not ruled out. For that reason, I henceforth assume that the kinds of
facts that the kind-theoretic conception invokes are natural.
What is it for a kind of facts to be natural in the relevant sense? I have in mind
something like Lewis’ (1983) notion of naturalness. A natural kind of facts is a kind
that is non-artificial, non-gerrymandered, a kind whose members are genuinely
similar. I do not want to be too specific about the characterisation of naturalness,
because I take it that the kind-theoretic conception of <–fundamentality should be
compatible with various views about the notion, ideally with any coherent view.
However, let me emphasise that ‘natural’ in the relevant sense should not be equated
with ‘perfectly natural’ à la Lewis (1983) or with ‘structural’ à la Sider (2011). It
is understood by these philosophers that what is perfectly natural or structural is
fundamental. Yet, as I see things, room should be left for the possibility that highly
non-fundamental kinds of facts be natural. It should be possible to hold, for instance,
that the kind mental fact and the kind logically complex fact both count as natural in
the relevant sense even though they are not perfectly natural or structural.
In the next section, I propose an account of <–fundamentality that is in line
with the kind-theoretic conception of fundamentality, and I do the same in the
following section with = –fundamentality and absolute fundamentality.

5 A kind‑theoretic account of <–fundamentality

Taking clause (5) in Bennett’s (MFT) seriously, the following characterisations of

foundation and <–fundamentality suggest themselves:
(Foundation-1) K founds ℒ ≡ (A) members of ℒ are typically or normally
grounded in some members of K  , and (B) neither K nor ℒ contains both
grounded and ungrounded facts.
(<–Fundamentality-1) F is more fundamental than G ≡ for some K and ℒ
such that F ∈ K  , F ∉ ℒ, G ∈ ℒ and G ∉ K  , K founds ℒ.
Given my second objection to (MFT), one gets a more satisfactory characteri-
sation of foundation if one removes the phrase ‘typically or normally’ in (Foun-
dation-1). If we do so, then condition (B) can be simplified since the condition
that ℒ does not contain both grounded and ungrounded facts is automatically
satisfied. The resulting characterisation thus goes as follows:
(Foundation-2) K founds ℒ ≡ (A) every member of ℒ is grounded in
some members of K  , and (B) K does not contain both grounded and
ungrounded facts.
Of course, the resulting characterisation of <–fundamentality is subject to my
objection against the view that (MFT)’s clause (5) is sufficient for a fact F to be

13
 Synthese

more fundamental than a fact G, namely objection (III). But let me for a moment
put <–fundamentality aside and focus first on the characterisation of foundation.

5.1 Foundation

Foundation, remember, is supposed to capture the idea that some kinds metaphysi-
cally arise from other kinds. This “generative” character of the relation makes
(Foundation-2) inadequate. One reason is the presence of condition (B). Bennett’s
argument for having condition (a) in (MFT)’s clause (5) turns out to be an argu-
ment for having the weaker condition that kind ℒ does not contain both grounded
and ungrounded facts; she does not argue that kind K should also be required to
satisfy this condition (Bennett, 2017, pp. 159–160). Be that as it may, having con-
dition (B) in the characterisation of foundation is objectionable. It should be pos-
sible to hold, say, both that the kind microphysical fact generates, in the intended
sense, the kind thermodynamical fact, and that the kind microphysical fact contains
both ungrounded members and grounded members (for instance conjunctions of
ungrounded microphysical facts).
Another reason to be dissatisfied with (Foundation-2) is that it does not guaran-
tee that foundation is irreflexive. Since foundation is a generative relation, it should
arguably have that property. Counterexamples to irreflexivity are actually not hard
to find. Suppose that that every member of the kind fact about physical entities is
grounded in some members of that same kind.17 (Foundation-2) then predicts that
the kind physical fact founds itself. Note that this counterexample also applies to the
characterisation obtained from (Foundation-2) by dropping condition (B).
In the light of these objections, it is natural to suggest the following alternative
characterisation of foundation:
(Foundation-3) K founds ℒ ≡ (A) every member of ℒ is grounded in some
members of K  , and (B) K ≠ ℒ.
But this suggestion will not do either. Granted that every member of the kind ther-
modynamical fact is grounded in some members of the kind microphysical fact,
(Foundation-3) predicts that the kind thermodynamical fact is founded in the kind
physical fact. This sounds bad given the generative character of foundation: if kind
ℒ arises from kind K  , then surely ℒ cannot be a subkind of K .
What about, then, replacing (B) by the condition that and ℒ do not overlap, i.e.,
that no fact belongs to both K and ℒ ? The resulting characterisation, namely
(Foundation-4) K founds ℒ ≡ (A) every member of ℒ is grounded in some
members of K  , and (B) K and ℒ do not overlap,
escapes all the previous objections. Yet it is still problematic, because it does
not guarantee that foundation is asymmetric. Given that foundation is a genera-
tive relation, it should arguably have that property. Here is a counterexample to

17
  The assumption may be false of our world, but it cannot be discarded on a priori grounds (see Rosen,
2010: p. 116).

13
Synthese

asymmetry. Suppose the kind mental fact comprises infinitely many subkinds
­M1, ­M2, …, that likewise kind physical fact comprises infinitely many subkinds
­P1, ­P2, …, that the kind mental fact does not overlap the kind physical fact, and
finally that for any positive integer n, every member of M ­ n is grounded in some
member of ­Pn and every member of P ­ n is grounded in some members of ­Mn+1.
This assumption is exotic, for sure, but it does not seem incoherent. Given this
assumption, (Foundation-4) predicts that the kind mental fact both founds and
is founded in the kind physical fact. Note that since these kinds are distinct, the
objection also affects (Foundation-3).
In order to escape this objection, I suggest that we strengthen the defining con-
dition in (Foundation-4) by requiring that no member of ℒ helps ground some
member of K  . The resulting characterisation, namely
(Foundation) K founds ℒ ≡ (A) every member of ℒ is grounded in some
members of K  , (B) no member of K is partially grounded in some mem-
ber of ℒ, and (C) K and ℒ do not overlap,
is my official proposal. So characterised, foundation is both transitive and irre-
flexive, and is therefore asymmetric. That it is transitive is established in the fol-
lowing proof:
Suppose that K founds ℒ and that ℒ founds ℳ. (i) By clause (A) in
(Foundation), it follows that every member of ℳ is grounded in some mem-
bers of ℒ and every member of ℒ is grounded in some members ofK  .
Since grounding obeys the cut principle, it follows that every member of
ℳ is grounded in some members of K  . (ii) Suppose for reductio that some
member of K is partially grounded in some member of ℳ. Since ℒ founds
ℳ, by clause (A) in (Foundation), every member of ℳ is partially grounded
in some member of ℒ. Since partial grounding is transitive, it follows that
some member of K is partially grounded in some member of ℒ. But given
the assumption that K founds ℒ and clause (B) in (Foundation), this is
impossible. Hence, no member of K is partially grounded in some member
of ℳ. (iii) Suppose for reductio that K overlaps ℳ. Let then F be in both
K and ℳ. Since ℒ founds ℳ, by clause (A) in (Foundation) and the fact
that F is in ℳ, F is partially grounded in some member of ℒ. But given the
assumption that K founds ℒ, clause (B) in (Foundation) and the fact that F
is in K  , this is impossible. Hence, K does not overlap ℳ.
Irreflexivity clearly follows from (C), but also from (A) and (B) taken jointly.
The presence of both (B) and (C) does not create redundancy, though, because
having just (A) & (B) or just (A) & (C) would yield problems. Keeping just (A)
& (C) means opting for (Foundation-4), and we saw that (Foundation-4) does not
secure the asymmetry of foundation. Keeping just (A) & (B) is problematic for
another reason. Given that a case of foundation is a case where one kind of facts
generates another kind of facts, it sounds intuitively correct to say that if a kind
K founds a kind ℒ, then K and ℒ do not overlap. Dropping condition (C) in
(Foundation) yields a characterisation that fails to guarantee that this is the case.

13
 Synthese

For let K0 be the kind ungrounded fact, K1 the kind fact grounded in facts in K0 ,
and not partially grounded in facts outside of K0 , and K2 the kind fact grounded
in facts in K1 . Consider then the kind K that is the union of K0 and K1 and the
kind ℒ that is the union of K1 and K2 (I suppose for the sake of the argument
that these are natural kinds). The characterisation under consideration predicts
that K founds ℒ, even though K and ℒ overlap.

5.2 <–fundamentality

Let me now turn to the issue of characterising <–fundamentality. Remember the

characterisation put forward at the beginning of Sect. 5:
(<–Fundamentality-1) F is more fundamental than G ≡ for some K and ℒ
such that F ∈ K  , F ∉ ℒ, G ∈ ℒ and G ∉ K  , K founds ℒ.
Given condition (C) in (Foundation), it can be simplified a bit:
(<–Fundamentality-2) F is more fundamental than G ≡ for some K and ℒ
such that F ∈ K and G ∈ ℒ, K founds ℒ.
This characterisation of <–fundamentality is very natural and simple, but it must be
rejected because it does not guarantee that the relation has the right formal proper-
ties (remember that I impose as a constraint on any account of <–fundamentality
that it should guarantee that the relation is both transitive an irreflexive, and hence
asymmetric—see the end of Sect. 2). With (Foundation) in place, (<–Fundamental-
ity-2) is subject to my objection against clause (5) in Bennett’s (MFT), i.e. objection
(III). A slightly simplified version of the objection can actually be raised.18 Replace
the kind conjunction of atomic facts by the kind atomic fact and the kind disjunc-
tion of conjunctions of atomic facts by the kind conjunction of atomic facts. Then
consider the plausible view that the kind atomic fact founds the kind conjunction
of atomic facts and the other plausible view that the kind biological fact founds the
kind mental fact. Then (<–Fundamentality-2) predicts that every conjunction of
atomic biological facts is both more and less fundamental than every atomic mental
fact.
The previous objection is to the effect that the asymmetry of <–fundamentality
is not guaranteed by the proposed characterisation. Here is a further objection,
this one to the effect that the proposed characterisation does not secure transitiv-
ity. Let K be the kind human biological fact, ℒ the kind human mental fact, ℒ*
the kind mental fact, and ℳ the kind social fact. The view that K founds ℒ and
ℒ* founds ℳ is plausible. Assume the view to be correct and let F be in K  , G
be in ℒ and hence in ℒ*, and H be in ℳ. Assuming (Foundation), (<–Funda-
mentality-2) predicts that F is more fundamental than G and G is more funda-
mental than H. Do we have that F is more fundamental than H? Granted that H is
a social fact that is not grounded in human mental facts but is rather a non-human

18
  This simplified version of the objection does not seriously threaten (MFT)’s clause (5), because it is
implausible to hold that the kind atomic fact does not contain both grounded and ungrounded members.

13
Synthese

social fact grounded in mental facts of a very different sort, the reply is plausibly
negative.
In the light of these considerations, a radical move suggests itself: assume that
distinct kinds of facts do not overlap. Given this assumption, the proposed objec-
tions against (<–Fundamentality-2) are ineffective since they crucially involve
overlapping kinds. It can actually be proved that given the assumption, (<–Fun-
damentality-2) characterises a relation that is both transitive and irreflexive, and
hence asymmetric. That the relation is irreflexive is easy to see. Establishing tran-
sitivity requires only a little bit of work:
Suppose F is more fundamental than G and G is more fundamental than H.
By (<–Fundamentality-2), there are kinds of facts K  , ℒ, ℒ* and ℳ such
that F ∈ K  , G ∈ ℒ, G ∈ ℒ*, H ∈ ℳ, K founds ℒ and ℒ* founds ℳ.
Given that distinct facts do not overlap, we then have ℒ = ℒ*. Since, as we
saw, foundation is transitive, it follows that K founds ℳ. Since F ∈ K and
H ∈ ℳ, we can conclude via (<–Fundamentality-2) that F is more funda-
mental than H.
However, the no-overlap assumption is problematic. Even if we focus, as I
argued we should, on natural kinds of facts, it is clear that the no-overlap assump-
tion is false: we should recognise that there are, say, both the kind physical fact
and the kind microphysical fact. It might be replied that we should relativise the
relation of <–fundamentality to categorisations, where a categorisation is defined
as a set of non-overlapping kinds. On that account, <–fundamentality would in
effect be a 3-place relation rather than a 2-place relation. However, the suggested
relativised relation is artificial, because the grouping of kinds into categorisa-
tions, that is, into sets of non-overlapping kinds, has no metaphysical significance
(in the present context at least). Be that as it may, the initial task was to find an
acceptable characterisation of a 2-place relation of being more fundamental than
between facts.
At this stage, it is tempting to try to tinker with the right-hand-side of (<–Fun-
damentality-2), perhaps by invoking universal quantifiers instead of existential
quantifiers over kinds, perhaps a mix of both, or by invoking some conditions on
kinds, or something that involves both sorts of moves. However, I wish to stay as
close as possible to the original Bennettian characterisation, and I accordingly
opt for a strategy of “building on top of” (<–Fundamentality-2). Let me use ‘F is
quasi-prior to G’ for the condition expressed by the right-hand-side of (<–Fun-
damentality-2). I argued in effect that one cannot identify <–fundamentality with
quasi-priority by giving counterexamples to both the asymmetry and the tran-
sitivity of quasi-priority. My proposal is to identify <–fundamentality not with
quasi-priority, but with the asymmetric closure of its transitive closure.
Let me spell this out. Where n is a positive integer, say that there is an n-chain
from F to G iff there are facts ­F0, …, ­Fn such that ­F0 = F, ­Fn = G and for all inte-
gers k such that 0 ≤ k ≤ n–1, ­Fk is quasi-prior to ­Fk+1. Then say that F is chained
to G iff there is an n-chain from F to G for some positive integer n. Chaining is
the transitive closure of quasi-priority. The proposal is to define <–fundamental-
ity as the asymmetric closure of chaining, that is, to define it as follows:

13
 Synthese

(<–Fundamentality) F is more fundamental than G ≡ F is chained to G and

G is not chained to F.
The transitive closure of any relation is transitive. The asymmetric closure of any
relation is asymmetric. The asymmetric closure of any transitive relation is itself
transitive. Given these three facts, <–fundamentality as characterised above is
both asymmetric and transitive (and of course, being asymmetric it is also irre-
flexive). As a consequence, (<–Fundamentality) is immune from the objections I
raised against (<–Fundamentality-2). Note that on the assumption that the kinds
of facts are disjoint (an assumption I argued against above), (<–Fundamentality)
is equivalent to (<–Fundamentality-2): on that assumption, as we saw, quasi-pri-
ority is both transitive and irreflexive; being transitive, it is identical to its transi-
tive closure, namely chaining; and being transitive and irreflexive, it is asymmet-
ric, and as a result it is identical to its asymmetric closure.
The proposed account of <–fundamentality has been designed to fit well with
the kind-theoretic conception of fundamentality as described in Sect. 4 and I take
it that it does justice to a number of judgments of <–fundamentality that many are
inclined to make under that conception. But it arguably does not do justice to all
such judgments. Consider for instance the view that the biological facts are more
fundamental than the mental facts. I guess that many would take it to be correct
granted a kind-theoretic conception of fundamentality. However, material that I
used against (<–Fundamentality-2) can be used to argue that the view is false
given (<–Fundamentality): if F is a conjunction of atomic biological facts and
G an atomic mental fact, and if the kind atomic fact founds the kind conjunction
of atomic facts and the kind biological fact founds the kind mental fact, then by
(<–Fundamentality) F is not more fundamental than G. Is this a problem?
I do not think so. When we look at the particular example, we see that two
forces pull in opposite directions, as it were: the logical features of the facts pull
in the direction of G being more fundamental than F, and the “subject matter” of
the facts pulls in the direction of F being more fundamental than G. By (<–Fun-
damentality), these two forces cancel each other, and we get that neither fact is
more fundamental than the other. This intuitively makes sense independently
from the proposed account: why should the subject matter of the facts prevail
over their logical features, or vice versa? Agreed, (<–Fundamentality) may not do
justice to the view that the biological facts are more fundamental than the mental
facts; but we have a natural account of why the view is incorrect.
Also, it is not clear that what I described as the inclination to endorse the view
that the biological facts are more fundamental than the mental facts is the inclina-
tion to endorse that very view. There are several other views in the vicinity which,
given plausible ground-theoretic assumptions about physical facts and biologi-
cal facts, mesh perfectly well with the proposed account of <–fundamentality, for
instance:

• The kind biological fact founds the kind mental fact;

• Every logically simple determinate physical fact is more fundamental than
every logically simple determinate biological fact;

13
Synthese

• Every logically simple determinate physical fact is more fundamental than every
biological fact whatsoever;
• Every biological fact is less fundamental than some physical fact, whereas no
biological fact is more fundamental than some physical fact.

It may well be that, in some cases at least, the inclination in question is that of
endorsing some such view.

6 Characterising = –fundamentality and absolute fundamentality

Given the proposed characterisation of <–fundamentality, how are = –fundamental-

ity and absolute fundamentality to be characterised? For = –fundamentality, I will
not be very original. The Litland-Bennett proposal, labelled ‘(AFA)’ in Sect.  3,
strikes me as adequate:
(=–Fundamentality) F is as fundamental as G ≡ for all facts F*, (i) F* is more
fundamental than F iff F* is more fundamental than G, and (ii) F is more fun-
damental than F* iff G is more fundamental than F*.
(Note that the proposal guarantees that = –fundamentality is an equivalence relation,
and hence satisfies the constraint on accounts of that relation that I formulated at the
end of Sect. 2.) The case of absolute fundamentality deserves more discussion. The
natural proposal, I take it, is to deem a fact absolutely fundamental when it is mini-
mal for the relation of being more fundamental than:
(Fundamentality) F is fundamental ≡ no fact is more fundamental than F.
Bennett, remember, has another characterisation:
(F) F is fundamental ≡ F is ungrounded.
It turns out that given (MFT), these two characterisations are provably equivalent.
For suppose F is ungrounded. A quick inspection of (MFT) reveals that on (MFT), if
a fact is less fundamental than another fact, then it is grounded. Consequently, given
(MFT), F is minimal for <–fundamentality. Conversely, suppose that F is grounded.
Then F is partially grounded, and by clause (2) of (MFT) it follows that some fact is
more fundamental than F, i.e., that F is not minimal for <–fundamentality.
By contrast, on my characterisation of <–fundamentality, (Fundamentality) and
(F) are not provably equivalent. Given (<–fundamentality), being ungrounded
provably implies being minimal for <–fundamentality: if a fact F is not minimal
for <–fundamentality, then some fact is quasi-prior to it, and this requires that F
belong to a kind whose members are all grounded. But being minimal for <–funda-
mentality does not provably imply being ungrounded. Finding convincing examples
to illustrate it is not easy. Here is a tentative suggestion: the fact that God exists. As
accompanying assumptions that would guarantee that this fact fits the bill, I suggest
the following two: (i) the fact that God exists is grounded in the fact that it is part

13
 Synthese

of His nature that He exists, and (ii) the fact that God exists only belongs to kinds
of facts that comprise at least some ungrounded members and hence that are not
founded in any other kinds of facts. Assumption (i) is reminiscent of (some versions)
of the ontological argument for the existence of God.19 In order to secure (ii), more
assumptions need to be made. The fact that God exists belongs to the kind existence
fact. We should accordingly assume that for some object x, the fact that x exists is
ungrounded. We might for instance assume that the empty set is such an object. The
fact that God exists belongs to the kind divine fact. We should accordingly assume
that some such fact is ungrounded. We might for instance assume that the fact that it
is part of God’s nature that He exists (the fact mentioned as a ground in assumption
(i) above) is a fact of that sort. The hope is that whatever kind is invoked, we will
always be able to secure, perhaps via extra assumptions, that the kind in question
comprises ungrounded facts.
Some might argue that it is a bad feature of my characterisation of absolute fun-
damentality that it leaves room for there being fundamental yet grounded facts. But I
disagree. I have nothing against the view that on some conceptions of fundamental-
ity, absolute fundamentality should be equated with ungroundedness. Yet through-
out this paper I have been concerned with one particular conception of fundamental-
ity, the kind-theoretic conception; and once this conception is taken for granted, I do
not see any problem with the idea that some fundamental facts may be grounded.20

Acknowledgements  I am grateful to the audience of an eidos meeting for useful comments on an ances-
tor of this paper and to Maria Scarpati for detailed comments on another version. Work on the paper was
supported by the Swiss National Science Foundation (project BSCGI0_157792).

Funding  Open Access funding provided by Université de Genève.

Open Access  This article is licensed under a Creative Commons Attribution 4.0 International License,
which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as
you give appropriate credit to the original author(s) and the source, provide a link to the Creative Com-
mons licence, and indicate if changes were made. The images or other third party material in this article
are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is
not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission
directly from the copyright holder. To view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​
ses/​by/4.​0/.

19
  But see Glazier (2017) against this kind of grounding claim.
20
 My account of absolute fundamentality is not the only account of the notion in ground-theoretic
terms that leaves room for the possibility of fundamental yet grounded facts. See Giannotti forthcoming
and Correia forthcoming for illustrations. Raven’s (2016) notion of fundamentality as ineliminability is
another illustration.

13
Synthese

References
Audi, P. (2012). Grounding: Toward a theory of the in-virtue-of relation. The Journal of Philosophy,
109(12), 685–711
Bennett, K. (2011). By our bootstraps. Philosophical Perspectives, 25, 27–41
Bennett, K. (2017). Making things up. Oxford University Press.
Bennett, K. (2019). Response to Leuenberger, Shumener and Thompson. Analysis, 79(2), 327–340
Correia, F. (2010). Grounding and truth-functions. Logique et Analyse, 53(211), 251–279
Correia, F. (2021). The logic of relative fundamentality. Synthese, 198(special issue), 1279–1301
Correia, F. forthcoming. Fundamentality from grounding trees, Synthese.
Dasgupta, S. (2014). On the plurality of grounds. Philosophers’ Imprint, 14(20), 1–28
deRosset, L. (2013). Grounding explanations. Philosophers’ Imprint, 13(7), 1–26
Dixon, S. T. (2016). What is the well-foundedness of grounding? Mind, 125(498), 439–468
Fine, K. (2012). Guide to ground. In F. Correia & B. Schnieder (Eds.), Metaphysical grounding: Under-
standing the structure of reality. (pp. 37–80). Cambridge University Press.
Glazier, M. (2017). Essentialist explanation. Philosophical Studies, 174(11), 2871–2889
Giannotti, J. forthcoming. Fundamental yet grounded, Theoria.
Kim, J. (2002). The layered model: Metaphysical considerations. Philosophical Explorations, 5(1), 2–20
Leuenberger, S. (2020). The fundamental: Ungrounded or all-grounding? Philosophical Studies, 177,
2647–2669
Lewis, D. (1983). New Work for a theory of universals. Australasian Journal of Philosophy, 61, 343–377
Litland, J. (2013). On some counterexamples to the transitivity of grounding. Essays in Philosophy,
14(1), 19–32
Litland, J. (2016). Pure logic of many-many ground. Journal of Philosophical Logic, 45(5), 531–577
Millikan, R. (1989). Biosemantics. The Journal of Philosophy, 86, 281–297
Rabin, G. O. (2018). Grounding orthodoxy and the layered conception. In G. Priest & R. L. Bliss (Eds.),
Reality and its structure. (pp. 37–49). Oxford University Press.
Raven, M. J. (2016). Fundamentality without foundations. Philosophy and Phenomenological Research,
93(3), 607–626
Rosen, G. (2010). Metaphysical dependence: grounding and reduction. In B. Hale & A. Hoffmann (Eds.),
Modality: Metaphysics, logic, and epistemology. (pp. 109–136). Oxford University Press.
Schaffer, J. (2009). On what grounds what. In D. Manley, D. J. Chalmers, & R. Wasserman (Eds.), Meta-
metaphysics: New essays on the foundations of ontology. (pp. 347–383). Oxford University Press.
Schaffer, J. (2012). Grounding, transitivity, and contrastivity. In F. Correia & B. Schnieder (Eds.), Meta-
physical grounding: Understanding the structure of reality. (pp. 122–138). Cambridge University
Press.
Shumener, E. (2017). The metaphysics of identity: Is identity fundamental? Philosophy Compass, 12(1),
1–13
Shumener, E. (2019). Building and surveying: Relative fundamentality in Karen Bennett’s Making things
up. Analysis, 79(2), 303–314
Sider, T. (2011). Writing the book of the world. Oxford University Press.
Wallner, M. (2021). The ground of ground, essence, and explanation. Synthese, 198(special issue),
1257–1277
Werner, J. forthcoming. A grounding-based measure of relative fundamentality, Synthese.

Publisher’s Note  Springer Nature remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.

13